I started this project because I wanted to see how far WordPress search could be pushed when performance is the primary goal.

I initially tried to build the fastest possible solution while staying within the constraints and standards required for a plugin distributed through the official WordPress Plugin Directory. However, those requirements also limit how aggressively the search architecture can be optimized.

For the WordPress.org version, I therefore had to work within the standard WordPress environment and APIs. While I optimized the search as much as possible, there is an inherent performance cost to going through the WordPress application layer and its APIs.

This led me to build a separate Pro architecture where I could remove those constraints. Instead of relying on the WordPress API for every search operation, the search engine can access the index directly and avoid loading the full WordPress application stack for each query. The difference in performance is substantial compared with searching through the standard WordPress API.

The other key part of the architecture is the NLP-based inverted index. Content is processed and indexed in advance, so a search query does not need to scan the entire collection of posts or products. The index is built with NLP processing in mind and currently supports 36 languages.

This approach makes it possible to perform very fast searches even on sites containing 100,000+ posts or products, where conventional WordPress search approaches can become increasingly expensive.

The free version of the project is available in the WordPress Plugin Directory:

https://wordpress.org/plugins/prismsearch/

The project website is:

https://prismsearch.io/

I'm interested in the technical side of this problem and would be curious to hear how others have approached search at scale in WordPress or other CMS platforms.